Seventeen visits last week, and not one from the market we sell to

The column I had not been reading

Our own site, nightfill.co, for the seven days to 23 September 2026: 17 sessions, 17 users. Every single one landed on the home page. Not one landed on a blog post. Zero clicks on a call to action, zero forms submitted.

That is a small number and I am not going to dress it up. But the size was not what stopped me. It was the country column.

• United States: 13 sessions
• Ireland: 2
• Canada: 1
• Thailand: 1

Mexico: zero. Colombia: zero. We write for independent hostels in Mexico and Colombia. Half the site is in Spanish. Last week it was read entirely outside the market it was built for.

I had been looking at the sessions number for two weeks and treating it as a volume problem. It is not a volume problem. Seventeen of the right people would be a good week. Seventeen of the wrong ones is a different failure with a different fix, and the only thing separating those two diagnoses is a column I was not opening.

Traffic has a source too, and it says who already knew your name

The same seven days, by source:

• Direct, no referrer: 11 sessions
• Bing, organic search: 4
• Facebook referral: 2
• Google, organic search: 0

Direct means the person typed the address or clicked a link that carried no referrer. It is not discovery. It is someone who already had the name, usually because we put it in front of them somewhere else. So 11 of 17 visits were the tail of work done off the site, 4 came from the search engine with the smaller share of this market, and search on Google brought nobody at all.

Eleven, four, two, zero. Four numbers, and between them they say the site is not yet a channel. It is a landing pad for traffic created elsewhere. That is a perfectly reasonable stage to be at. It is a terrible thing to misread as an SEO problem and spend three months writing against.

What an average position of 5 is worth when it is measured over one impression

Search Console, 17 to 23 September, the whole week: four pages got impressions. Five impressions in total. Zero clicks. The home page: 2 impressions, average position 8.5. One blog post, English and Spanish: 1 impression each, average position 5. One more Spanish post: 1 impression, position 9. The query breakdown came back empty.

Average position 5 reads like a result. It is one person, one search, once. If I had put that number on a slide with the word average in front of it, it would have looked like ranking. It is a sample of one with a decimal point.

The discipline that costs nothing: every time a rate or a position or an average shows up on a dashboard, find the denominator before you react to it. Five impressions does not support a conclusion about ranking. It supports exactly one conclusion, which is that almost nobody is searching for what those pages are about, in the places we are being shown.

The same three columns exist in your analytics

Country, source, landing page. Most 20 to 80 bed properties I talk to look at one number, sessions, or do not look at all. Those three columns take four minutes to pull and they answer a question that decides where your next month of effort goes.

Here is the check. Pull the country report for your own site for the last 90 days and write down the top five. Then pull nationality at check-in from your PMS for the same 90 days and write down the top five there. Your property already records that at the front desk; the report exists whether or not anyone has ever opened it.

Then compare the two lists. There are two ways they come apart and they mean opposite things.

Your site is domestic and your beds are foreign

The OTA is doing all of your international reach. Every guest from a country that never appears in your site traffic is a guest the channel found for you, and you are paying the channel's price for exactly that service. Rewriting your home page will not change this. Being findable in those countries might, and that is a distribution problem, not a copy problem.

Your site matches your beds and still converts nobody

Then the right people are arriving and leaving, and the problem is on the page: the price, the form, the load time, the photos, the fact that the booking engine opens in a new tab and looks like a different company. That is a much cheaper problem to fix, and it is worth knowing which of the two you have before you start.

What the commission is actually buying

It is easy to write about direct bookings as though the OTA were simply taking money. On one real 8 bed-night Hostelworld booking at our own hostel in Oaxaca, the three promotions on it multiplied rather than added, 0.9 x 0.9 x 0.9 = 0.729, and the effective commission worked out at 26.7 percent. Our pricing engine carries those economics as constants rather than assumptions: Booking at a 1.20 markup against a 0.20 commission, Hostelworld at 1.27 against 0.209, and a hard floor of 130 MXN net per bed that no stack of promotions is allowed to break.

Twenty-six point seven percent is what reach costs when you have none of your own. That is the honest framing. The channel is selling you an audience in thirteen countries you have no way to appear in, and it is charging roughly a quarter of the booking for it. The work of building direct is the work of buying that audience back, one country at a time, and it only pays in the countries where you can actually be found today.

What we are changing, and what will show it worked

Not more posts. We are at six blog posts in the last seven days against a target of seven, and the volume is not the constraint. The constraint is that the Spanish pages are not in front of Spanish-speaking operators at all: zero sessions from Mexico or Colombia, and zero Google organic sessions, in the same week the Spanish pages were being published.

So the change is placement, not production. The Spanish pages go where operators in those two countries already are, rather than waiting to be found. I will judge it on the same column, 30 days from now: sessions from Mexico or Colombia above zero, and the Google organic row above zero. Both are currently zero, which makes it an easy measurement to be honest about.

Two caveats, because without them the numbers above say more than they know. Seventeen sessions across one week is far too small to say anything about relative interest between countries; what it does establish, and this needs no sample size at all, is that the market we sell to did not visit. And our site is a business-to-business site. Yours sells beds, so its shape will be different. The three columns are the same.

If this sounds like your property, see the three Nightfill tiers and start a free 14-day pilot on your own data: view pricing.